The following chunk is run once to initialize the pipeline, which

See _targets.R for more information.

# tar_make()

All analyses were run in R (R version 4.3.1 (2023-06-16)). Please note the session info for package version information.

## R version 4.3.1 (2023-06-16)
## Platform: x86_64-apple-darwin20 (64-bit)
## Running under: macOS Sonoma 14.2
## 
## Matrix products: default
## BLAS:   /Library/Frameworks/R.framework/Versions/4.3-x86_64/Resources/lib/libRblas.0.dylib 
## LAPACK: /Library/Frameworks/R.framework/Versions/4.3-x86_64/Resources/lib/libRlapack.dylib;  LAPACK version 3.11.0
## 
## locale:
## [1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8
## 
## time zone: Europe/Berlin
## tzcode source: internal
## 
## attached base packages:
## [1] stats     graphics  grDevices utils     datasets  methods   base     
## 
## other attached packages:
##  [1] LDM_6.0               vegan_2.6-4           lattice_0.21-8       
##  [4] permute_0.9-7         phyloseq_1.44.0       ggbeeswarm_0.7.2     
##  [7] kableExtra_1.3.4.9000 gridExtra_2.3         cowplot_1.1.2        
## [10] ggpubr_0.6.0          lubridate_1.9.2       forcats_1.0.0        
## [13] stringr_1.5.0         dplyr_1.1.3           purrr_1.0.2          
## [16] readr_2.1.4           tidyr_1.3.0           tibble_3.2.1         
## [19] ggplot2_3.4.4         tidyverse_2.0.0       tarchetypes_0.7.8    
## [22] targets_1.2.2         knitr_1.43           
## 
## loaded via a namespace (and not attached):
##   [1] rstudioapi_0.15.0       jsonlite_1.8.8          magrittr_2.0.3         
##   [4] rmarkdown_2.24          fs_1.6.3                zlibbioc_1.46.0        
##   [7] vctrs_0.6.5             multtest_2.56.0         RCurl_1.98-1.12        
##  [10] rstatix_0.7.2           webshot_0.5.5           htmltools_0.5.7        
##  [13] broom_1.0.5             Rhdf5lib_1.22.0         rhdf5_2.44.0           
##  [16] sass_0.4.8              parallelly_1.36.0       bslib_0.6.1            
##  [19] naturalsort_0.1.3       plyr_1.8.8              cachem_1.0.8           
##  [22] igraph_1.5.1            lifecycle_1.0.4         iterators_1.0.14       
##  [25] pkgconfig_2.0.3         Matrix_1.6-1            R6_2.5.1               
##  [28] fastmap_1.1.1           clue_0.3-64             GenomeInfoDbData_1.2.10
##  [31] future_1.33.0           digest_0.6.34           colorspace_2.1-0       
##  [34] spatial_7.3-17          furrr_0.3.1             S4Vectors_0.38.1       
##  [37] ps_1.7.5                RSpectra_0.16-1         base64url_1.4          
##  [40] fansi_1.0.6             timechange_0.2.0        httr_1.4.7             
##  [43] abind_1.4-5             mgcv_1.9-0              compiler_4.3.1         
##  [46] withr_3.0.0             backports_1.4.1         BiocParallel_1.34.2    
##  [49] carData_3.0-5           ggsignif_0.6.4          MASS_7.3-60            
##  [52] castor_1.7.11           biomformat_1.28.0       fBasics_4022.94        
##  [55] tools_4.3.1             vipor_0.4.5             beeswarm_0.4.0         
##  [58] ape_5.7-1               quadprog_1.5-8          glue_1.7.0             
##  [61] stabledist_0.7-1        callr_3.7.3             nlme_3.1-163           
##  [64] rhdf5filters_1.12.1     grid_4.3.1              cluster_2.1.4          
##  [67] reshape2_1.4.4          ade4_1.7-22             generics_0.1.3         
##  [70] gtable_0.3.4            tzdb_0.4.0              data.table_1.14.8      
##  [73] hms_1.1.3               xml2_1.3.5              car_3.1-2              
##  [76] utf8_1.2.4              XVector_0.40.0          rmutil_1.1.10          
##  [79] BiocGenerics_0.46.0     ggrepel_0.9.3           foreach_1.5.2          
##  [82] pillar_1.9.0            splines_4.3.1           survival_3.5-7         
##  [85] GUniFrac_1.7            tidyselect_1.2.0        Biostrings_2.68.1      
##  [88] future.callr_0.8.2      IRanges_2.34.1          svglite_2.1.1          
##  [91] stats4_4.3.1            xfun_0.40               Biobase_2.60.0         
##  [94] statmod_1.5.0           timeDate_4022.108       matrixStats_1.0.0      
##  [97] stringi_1.7.12          yaml_2.3.7              evaluate_0.21          
## [100] codetools_0.2-19        timeSeries_4031.107     cli_3.6.2              
## [103] rpart_4.1.19            systemfonts_1.0.4       munsell_0.5.0          
## [106] processx_3.8.2          jquerylib_0.1.4         Rcpp_1.0.12            
## [109] GenomeInfoDb_1.36.3     stable_1.1.6            globals_0.16.2         
## [112] modeest_2.4.0           parallel_4.3.1          bitops_1.0-7           
## [115] phangorn_2.11.1         listenv_0.9.0           viridisLite_0.4.2      
## [118] scales_1.3.0            statip_0.2.3            crayon_1.5.2           
## [121] rlang_1.1.3             fastmatch_1.1-4         rvest_1.0.3

Manuscript

Table 1. Baseline Characteristics of Participants.

NC MCI p p test
n 200 58
Gender (Female/Male) 87/113 15/43 0.022 fisher
Years of Education (0-10/11-16/16+) 24/110/66 16/24/18 0.018 fisher
First Language (Other/French/Luxembourgish/German) 18/182 6/52 0.798 fisher
Living With Partner (No/Yes) 61/139 17/41 1.000 fisher
BDI I mild (Yes/No) 23/177 9/49 0.497 fisher
APOE4 (At Least 1/None) 54/146 19/39 0.410 fisher
ATB in last 6 months (YES/NO) 21/179 9/49 0.351 fisher
Age 63.76 ± 7.84 67.6 ± 9 0.004 t
BMI 27.04 ± 4.37 29.74 ± 5.91 0.002 t
Chao1 311.56 ± 64.96 295.05 ± 86.37 0.181 t
Shannon 3.99 ± 0.38 3.93 ± 0.46 0.415 t
InvSimpson 27.75 ± 11.46 26.79 ± 11.66 0.580 t

Note. T = Student’s t-Test, Fisher = Fisher’s Exact Test. Numbers refer to means ± standard deviations for continuous, n for categorical characteristics. BMI = body mass index, BDI-I = Beck Depression Inventory I, APOE = apolipoprotein ε4 status.

## Mean age (SD) age of participants = 64.6 ( 8.3 ) years

Table 2. Taxonomic Analysis Across Groups of Education (DESeq2).

Basic Adjustments

Level Taxon low.high low.int
Class Actinobacteria 0.035
Class Bacilli 0.008
Family Streptococcaceae 0.018
Genus Streptococcus 0.048
Order Lactobacillales 0.009
a Shown results are restricted to p < .05.

Adjusted for age, sex/gender, BMI, and use of antibiotic medication in the last 6 months.

Basic + Further Adjustments

Level Taxon low.high low.int
Class Bacilli 0.009
Family Streptococcaceae 0.029
Order Lactobacillales 0.016
a Shown results are restricted to p < .05.

Adjusted for age, sex/gender, BMI, use of antibiotic medication in the last 6 months, BDI-I, first language, and partnership status.

Basic + Further Adjustments + APOE

Level Taxon low.high low.int
Class Bacilli 0.01
Family Streptococcaceae 0.044
Order Lactobacillales 0.024
a Shown results are restricted to p < .05.

Adjusted for age, sex/gender, BMI, use of antibiotic medication in the last 6 months, BDI-I, first language, partnership status and APOE.

Table 2. Taxonomic Analysis Across Groups of Education (ANCOM-BC).

Basic Adjustments

Level Taxon low.high
Genus Lachnospiraceae_UCG_001 0.009
ASV OTU_000053 0.048
ASV OTU_000508 0.048
a Shown results are restricted to p < .05.

Adjusted for age, sex/gender, BMI, and use of antibiotic medication in the last 6 months.

Basic + Further Adjustments

Level Taxon low.high
Genus Lachnospiraceae_UCG_001 0.019
a Shown results are restricted to p < .05.

Adjusted for age, sex/gender, BMI, use of antibiotic medication in the last 6 months, BDI-I, first language, and partnership status.

Basic + Further Adjustments + APOE

Level Taxon low.high
Genus Lachnospiraceae_UCG_001 0.021
a Shown results are restricted to p < .05.

Adjusted for age, sex/gender, BMI, use of antibiotic medication in the last 6 months, BDI-I, first language, partnership status and APOE.

Table 3. Mediation Analysis with Chao1 as Mediator.

Note. Results of mediation analysis with interaction terms of education and Chao1 in the outcome model. Standard errors were estimated with 2000 bootstraps. * P Value < .05, ** P Value < .01, *** P Value < .001. Rcde: controlled direct effect odds ratio (referring to CDE); Rpnde: pure natural direct effect odds ratio (referring to NDE); Rtnde: total natural direct effect odds ratio; Rpnie: pure natural indirect effect odds ratio; Rtnie: total natural indirect effect odds ratio (referring to NIE); Rte: total effect odds ratio; Ercde: excess relative risk due to controlled direct effect; Erintref: excess relative risk due to reference interaction; Erintmed: excess relative risk due to mediated interaction; Erpnie: excess relative risk due to pure natural indirect effect; Ercde(P): proportion Ercde; Erintref(P): proportion Erintref; Erintmed(P): proportion Erintmed; Erpnie(P): proportion Erpnie; pm: overall proportion mediated; int: overall proportion attributable to interaction; pe: overall proportion eliminated). Cells with – indicate n/a.

0-10 vs. 11-16

11-16 vs. 0-10 YEDU - Chao1 - no interaction
Effect Estimate [95% CI] P Value Sig
Rcde 0.34 [ 0.16 to 0.74] 0.010 ’*’
Rpnde 0.34 [ 0.16 to 0.74] 0.010 ’*’
Rtnde 0.34 [ 0.16 to 0.74] 0.010 ’*’
Rpnie 0.92 [ 0.77 to 1.07] 0.284
Rtnie 0.92 [ 0.76 to 1.07] 0.284
Rte 0.32 [ 0.15 to 0.68] 0.006 ’**’
pm 0.04 [-0.04 to 0.25] 0.288

0-10 vs. 16+

16+ vs. 0-10 YEDU - Chao1 - no interaction
Effect Estimate [95% CI] P Value Sig
Rcde 0.33 [ 0.15 to 0.76] 0.015 ’*’
Rpnde 0.34 [ 0.15 to 0.76] 0.015 ’*’
Rtnde 0.34 [ 0.15 to 0.76] 0.015 ’*’
Rpnie 0.93 [ 0.78 to 1.06] 0.316
Rtnie 0.93 [ 0.77 to 1.07] 0.316
Rte 0.31 [ 0.14 to 0.71] 0.008 ’**’
pm 0.03 [-0.04 to 0.24] 0.321

0-10 vs. 11-16 multinomial

11-16 vs. 0-10 YEDU - Chao1 - no interaction - multinomial
Effect Estimate [95% CI] P Value Sig
Rcde 0.47 [ 0.28 to 0.80] 0.010 ’*’
Rpnde 0.48 [ 0.29 to 0.80] 0.010 ’*’
Rtnde 0.47 [ 0.28 to 0.80] 0.010 ’*’
Rpnie 0.96 [ 0.85 to 1.04] 0.284
Rtnie 0.94 [ 0.81 to 1.06] 0.284
Rte 0.45 [ 0.27 to 0.75] 0.006 ’**’
pm 0.05 [-0.05 to 0.29] 0.288

0-10 vs. 16+ multinomial

16+ vs. 0-10 YEDU - Chao1 - no interaction - multinomial
Effect Estimate [95% CI] P Value Sig
Rcde 0.46 [ 0.26 to 0.82] 0.015 ’*’
Rpnde 0.47 [ 0.26 to 0.82] 0.015 ’*’
Rtnde 0.46 [ 0.26 to 0.82] 0.015 ’*’
Rpnie 0.96 [ 0.86 to 1.04] 0.316
Rtnie 0.94 [ 0.81 to 1.05] 0.316
Rte 0.44 [ 0.25 to 0.79] 0.008 ’**’
pm 0.05 [-0.04 to 0.28] 0.321

0-10 vs. 11-16 with interaction

11-16 vs. 0-10 YEDU - Chao1 - with interaction
Effect Estimate [95% CI] P Value Sig
Rcde 0.33 [ 0.15 to 0.76] 0.012 ’*’
Rpnde 0.33 [ 0.15 to 0.73] 0.012 ’*’
Rtnde 0.33 [ 0.15 to 0.77] 0.011 ’*’
Rpnie 0.95 [ 0.70 to 1.19] 0.647
Rtnie 0.94 [ 0.71 to 1.33] 0.693
Rte 0.31 [ 0.15 to 0.68] 0.006 ’**’
ERcde -0.53 [-0.74 to -0.17] 0.012 ’*’
ERintref -0.14 [-0.32 to 0.05] 0.114
ERintmed 0.03 [-0.23 to 0.31] 0.792
ERpnie -0.05 [-0.30 to 0.19] 0.647
ERcde(prop) 0.77 [ 0.42 to 0.99] 0.008 ’**’
ERintref(prop) 0.20 [-0.09 to 0.56] 0.115
ERintmed(prop) -0.04 [-0.52 to 0.41] 0.796
ERpnie(prop) 0.07 [-0.32 to 0.53] 0.651
pm 0.03 [-0.15 to 0.28] 0.694
int 0.16 [-0.05 to 0.40] 0.104
pe 0.23 [ 0.01 to 0.58] 0.043 ’*’

0-10 vs. 16+ with interaction

16+ vs. 0-10 YEDU - Chao1 - with interaction
Effect Estimate [95% CI] P Value Sig
Rcde 0.33 [ 0.14 to 0.78] 0.018 ’*’
Rpnde 0.35 [ 0.15 to 0.81] 0.016 ’*’
Rtnde 0.33 [ 0.15 to 0.79] 0.016 ’*’
Rpnie 0.95 [ 0.71 to 1.17] 0.662
Rtnie 0.89 [ 0.68 to 1.14] 0.334
Rte 0.31 [ 0.14 to 0.72] 0.008 ’**’
ERcde -0.53 [-0.76 to -0.14] 0.018 ’*’
ERintref -0.11 [-0.28 to 0.10] 0.218
ERintmed 0.01 [-0.25 to 0.25] 0.966
ERpnie -0.05 [-0.29 to 0.17] 0.662
ERcde(prop) 0.78 [ 0.41 to 1.00] 0.010 ’*’
ERintref(prop) 0.17 [-0.20 to 0.49] 0.216
ERintmed(prop) -0.01 [-0.41 to 0.45] 0.964
ERpnie(prop) 0.07 [-0.29 to 0.52] 0.668
pm 0.06 [-0.06 to 0.38] 0.340
int 0.16 [-0.02 to 0.38] 0.082
pe 0.22 [ 0.00 to 0.59] 0.049 ’*’

0-10 vs. 11-16 multinomial with interaction

11-16 vs. 0-10 YEDU - Chao1 - with interaction - multinomial
Effect Estimate [95% CI] P Value Sig
Rcde 0.46 [ 0.26 to 0.81] 0.012 ’*’
Rpnde 0.47 [ 0.26 to 0.80] 0.012 ’*’
Rtnde 0.46 [ 0.28 to 0.82] 0.011 ’*’
Rpnie 0.97 [ 0.80 to 1.10] 0.647
Rtnie 0.95 [ 0.76 to 1.27] 0.693
Rte 0.44 [ 0.27 to 0.75] 0.006 ’**’
ERcde -0.53 [-0.74 to -0.17] 0.012 ’*’
ERintref -0.01 [-0.20 to 0.16] 0.998
ERintmed 0.01 [-0.16 to 0.22] 0.891
ERpnie -0.03 [-0.20 to 0.10] 0.647
ERcde(prop) 0.95 [ 0.54 to 1.22] 0.008 ’**’
ERintref(prop) 0.01 [-0.36 to 0.45] 0.998
ERintmed(prop) -0.01 [-0.48 to 0.37] 0.894
ERpnie(prop) 0.05 [-0.23 to 0.46] 0.651
pm 0.04 [-0.21 to 0.33] 0.694
int 0.00 [-0.18 to 0.11] 0.677
pe 0.05 [-0.22 to 0.46] 0.765

0-10 vs. 16+ multinomial with interaction

16+ vs. 0-10 YEDU - Chao1 - with interaction - multinomial
Effect Estimate [95% CI] P Value Sig
Rcde 0.45 [ 0.24 to 0.83] 0.018 ’*’
Rpnde 0.49 [ 0.27 to 0.86] 0.016 ’*’
Rtnde 0.46 [ 0.26 to 0.84] 0.016 ’*’
Rpnie 0.97 [ 0.81 to 1.09] 0.662
Rtnie 0.91 [ 0.74 to 1.11] 0.334
Rte 0.44 [ 0.26 to 0.79] 0.008 ’**’
ERcde -0.53 [-0.76 to -0.14] 0.018 ’*’
ERintref 0.02 [-0.16 to 0.20] 0.757
ERintmed -0.02 [-0.18 to 0.16] 0.850
ERpnie -0.03 [-0.19 to 0.09] 0.662
ERcde(prop) 0.96 [ 0.53 to 1.25] 0.010 ’*’
ERintref(prop) -0.04 [-0.49 to 0.36] 0.758
ERintmed(prop) 0.03 [-0.33 to 0.44] 0.852
ERpnie(prop) 0.05 [-0.19 to 0.44] 0.668
pm 0.08 [-0.09 to 0.41] 0.340
int -0.01 [-0.18 to 0.12] 0.647
pe 0.04 [-0.25 to 0.47] 0.803

Mediation with LDM

## ------------------------------------------------------------------ 
## 
## LDM results 
## identified otus: 
## identified otus q value: 
## ldm-med global omni3 p = 0.9935
## bivariate association of exposure and microbiome: 0.0555
## bivariate association of microbiome and outcome conditional on exposure: 0.1498
## number of permutations completed: 48000
## stopping criterion met by all global tests: TRUE
## stopping criterion met by all OTU-specific tests: TRUE
## 
## ------------------------------------------------------------------ 
## 
## PermanovaFL results 
## permanova-med global omni p = 0.535
## permanova per jaccard p = 0.345
## permanova per bray p = 0.696666666666667
## number of permutations completed: 600
## stopping criterion met by all tests of covariates: TRUE
## 
## ------------------------------------------------------------------

Figure 1. Alpha Diversity Across Age and Education Groups.

Note. Panels show results stratified by age and education groups with 0-10, 11-16 and 16+ years of education. Reported P values refer to Student’s t-Tests. InvSimpson = Inverse Simpson. Author MK.

Figure 2. Ordination Plots for Education Groups.

A ANOVA

Df Sum Sq Mean Sq F value Pr(>F)
Groups 2 0.01 0 1.8 0.167
Residuals 255 0.63 0

B ANOVA

Df Sum Sq Mean Sq F value Pr(>F)
Groups 2 0.02 0.01 3.11 0.048
Residuals 126 0.32 0.00

Note. Ordination using Non-metric Multidimensional Scaling based on Bray-Curtis dissimilarity. P value (adonis2) adjusted for gender, age, ATB, BDI-I, first language, PS, and APOE. Ordination using Non-metric Multidimensional Scaling based on Bray-Curtis dissimilarity. A Full sample. B Restricted sample age 65 and older. Authors MK and VTEA.

Supplement

Table 1. Coefficients of Regression Models with Chao1.

Without interaction

Outcome Model
Estimate [95% CI] P Value Sig
(Intercept) -4.05 [-7.22 to -1.04] 0.010 ’*’
Years_of_Education11-16 -1.21 [-2.07 to -0.35] 0.006 ’**’
Years_of_Education16+ -1.23 [-2.16 to -0.31] 0.009 ’**’
zChao1 -0.21 [-0.54 to 0.11] 0.198
Age 0.07 [ 0.03 to 0.11] 0.001 ’**’
GenderMale 1.11 [ 0.37 to 1.91] 0.004 ’**’
ATB_in_last_6_monthsNO -0.47 [-1.36 to 0.49] 0.319
BDI_I_mildNo -0.05 [-0.95 to 0.94] 0.923
First_LanguageFrench/Luxembourgish/German -0.33 [-1.34 to 0.80] 0.545
Living_With_PartnerYes -0.11 [-0.83 to 0.63] 0.761
APOE4None -0.62 [-1.32 to 0.09] 0.084
Mediator Model
Estimate [95% CI] P Value Sig
(Intercept) -1.48 [-2.65 to -0.31] 0.014 ’*’
Years_of_Education11-16 0.42 [ 0.07 to 0.77] 0.018 ’*’
Years_of_Education16+ 0.38 [ 0.00 to 0.76] 0.050
Age 0.01 [-0.01 to 0.02] 0.258
GenderMale 0.06 [-0.21 to 0.32] 0.671
ATB_in_last_6_monthsNO 0.40 [ 0.02 to 0.77] 0.039 ’*’
BDI_I_mildNo 0.04 [-0.33 to 0.41] 0.833
First_LanguageFrench/Luxembourgish/German -0.15 [-0.58 to 0.27] 0.478
Living_With_PartnerYes 0.33 [ 0.07 to 0.60] 0.015 ’*’
APOE4None 0.08 [-0.19 to 0.35] 0.561

With interaction

Outcome Model
Estimate [95% CI] P Value Sig
(Intercept) -4.09 [-7.27 to -1.06] 0.010 ’*’
Years_of_Education11-16 -1.24 [-2.12 to -0.35] 0.005 ’**’
Years_of_Education16+ -1.26 [-2.22 to -0.30] 0.010 ’*’
zChao1 -0.14 [-0.76 to 0.44] 0.640
Age 0.07 [ 0.03 to 0.11] 0.002 ’**’
GenderMale 1.10 [ 0.35 to 1.91] 0.005 ’**’
ATB_in_last_6_monthsNO -0.47 [-1.37 to 0.48] 0.314
BDI_I_mildNo -0.07 [-0.98 to 0.92] 0.880
First_LanguageFrench/Luxembourgish/German -0.31 [-1.33 to 0.82] 0.572
Living_With_PartnerYes -0.10 [-0.82 to 0.65] 0.797
APOE4None -0.63 [-1.33 to 0.08] 0.080
Years_of_Education11-16:zChao1 -0.02 [-0.80 to 0.79] 0.967
Years_of_Education16+:zChao1 -0.20 [-1.00 to 0.62] 0.630
Mediator Model
Estimate [95% CI] P Value Sig
(Intercept) -1.48 [-2.65 to -0.31] 0.014 ’*’
Years_of_Education11-16 0.42 [ 0.07 to 0.77] 0.018 ’*’
Years_of_Education16+ 0.38 [ 0.00 to 0.76] 0.050
Age 0.01 [-0.01 to 0.02] 0.258
GenderMale 0.06 [-0.21 to 0.32] 0.671
ATB_in_last_6_monthsNO 0.40 [ 0.02 to 0.77] 0.039 ’*’
BDI_I_mildNo 0.04 [-0.33 to 0.41] 0.833
First_LanguageFrench/Luxembourgish/German -0.15 [-0.58 to 0.27] 0.478
Living_With_PartnerYes 0.33 [ 0.07 to 0.60] 0.015 ’*’
APOE4None 0.08 [-0.19 to 0.35] 0.561

Note. Regression coefficients for mediator and outcome models, used for mediation analysis with alpha diversity measures. All analyses were adjusted for age, sex/gender, ATB, BDI-I, first language, PS and APOE. * P Value < .05, ** P Value < .01, *** P Value < .001.

Table 2. Coefficients of Regression Models with Shannon.

Without interaction

Outcome Model
Estimate [95% CI] P Value Sig
(Intercept) -3.89 [-7.01 to -0.90] 0.012 ’*’
Years_of_Education11-16 -1.26 [-2.11 to -0.41] 0.003 ’**’
Years_of_Education16+ -1.25 [-2.18 to -0.34] 0.007 ’**’
zShannon -0.13 [-0.46 to 0.19] 0.423
Age 0.06 [ 0.03 to 0.11] 0.002 ’**’
GenderMale 1.10 [ 0.37 to 1.90] 0.005 ’**’
ATB_in_last_6_monthsNO -0.49 [-1.38 to 0.46] 0.289
BDI_I_mildNo -0.06 [-0.96 to 0.91] 0.893
First_LanguageFrench/Luxembourgish/German -0.31 [-1.33 to 0.81] 0.562
Living_With_PartnerYes -0.16 [-0.87 to 0.57] 0.666
APOE4None -0.63 [-1.33 to 0.07] 0.077
Mediator Model
Estimate [95% CI] P Value Sig
(Intercept) -1.62 [-2.79 to -0.44] 0.007 ’**’
Years_of_Education11-16 0.36 [ 0.01 to 0.71] 0.042 ’*’
Years_of_Education16+ 0.37 [-0.01 to 0.75] 0.055
Age 0.01 [ 0.00 to 0.03] 0.102
GenderMale 0.09 [-0.17 to 0.35] 0.494
ATB_in_last_6_monthsNO 0.43 [ 0.05 to 0.80] 0.026 ’*’
BDI_I_mildNo 0.05 [-0.32 to 0.42] 0.778
First_LanguageFrench/Luxembourgish/German -0.20 [-0.62 to 0.23] 0.364
Living_With_PartnerYes 0.26 [-0.01 to 0.53] 0.057
APOE4None 0.01 [-0.26 to 0.28] 0.938

With interaction

Outcome Model
Estimate [95% CI] P Value Sig
(Intercept) -4.10 [-7.27 to -1.08] 0.009 ’**’
Years_of_Education11-16 -1.31 [-2.18 to -0.45] 0.003 ’**’
Years_of_Education16+ -1.28 [-2.22 to -0.36] 0.007 ’**’
zShannon -0.01 [-0.61 to 0.59] 0.985
Age 0.07 [ 0.03 to 0.11] 0.001 ’**’
GenderMale 1.08 [ 0.34 to 1.87] 0.006 ’**’
ATB_in_last_6_monthsNO -0.51 [-1.40 to 0.45] 0.279
BDI_I_mildNo -0.11 [-1.02 to 0.88] 0.825
First_LanguageFrench/Luxembourgish/German -0.29 [-1.32 to 0.84] 0.591
Living_With_PartnerYes -0.14 [-0.85 to 0.59] 0.698
APOE4None -0.64 [-1.34 to 0.06] 0.072
Years_of_Education11-16:zShannon 0.04 [-0.77 to 0.87] 0.924
Years_of_Education16+:zShannon -0.38 [-1.19 to 0.43] 0.354
Mediator Model
Estimate [95% CI] P Value Sig
(Intercept) -1.62 [-2.79 to -0.44] 0.007 ’**’
Years_of_Education11-16 0.36 [ 0.01 to 0.71] 0.042 ’*’
Years_of_Education16+ 0.37 [-0.01 to 0.75] 0.055
Age 0.01 [ 0.00 to 0.03] 0.102
GenderMale 0.09 [-0.17 to 0.35] 0.494
ATB_in_last_6_monthsNO 0.43 [ 0.05 to 0.80] 0.026 ’*’
BDI_I_mildNo 0.05 [-0.32 to 0.42] 0.778
First_LanguageFrench/Luxembourgish/German -0.20 [-0.62 to 0.23] 0.364
Living_With_PartnerYes 0.26 [-0.01 to 0.53] 0.057
APOE4None 0.01 [-0.26 to 0.28] 0.938

Note. Regression coefficients for mediator and outcome models, used for mediation analysis with alpha diversity measures. All analyses were adjusted for age, sex/gender, ATB, BDI-I, first language, PS and APOE. * P Value < .05, ** P Value < .01, *** P Value < .001.

Table 3. Coefficients of Regression Models with Inverse Simpson.

Without interaction

Outcome Model
Estimate [95% CI] P Value Sig
(Intercept) -3.72 [-6.78 to -0.79] 0.014 ’*’
Years_of_Education11-16 -1.29 [-2.14 to -0.46] 0.002 ’**’
Years_of_Education16+ -1.28 [-2.21 to -0.38] 0.006 ’**’
zInvSimpson -0.08 [-0.42 to 0.25] 0.625
Age 0.06 [ 0.02 to 0.10] 0.002 ’**’
GenderMale 1.09 [ 0.36 to 1.89] 0.005 ’**’
ATB_in_last_6_monthsNO -0.53 [-1.41 to 0.42] 0.254
BDI_I_mildNo -0.08 [-0.97 to 0.88] 0.862
First_LanguageFrench/Luxembourgish/German -0.31 [-1.32 to 0.82] 0.567
Living_With_PartnerYes -0.18 [-0.88 to 0.55] 0.629
APOE4None -0.63 [-1.33 to 0.08] 0.077
Mediator Model
Estimate [95% CI] P Value Sig
(Intercept) -0.88 [-2.08 to 0.31] 0.147
Years_of_Education11-16 0.17 [-0.18 to 0.52] 0.342
Years_of_Education16+ 0.22 [-0.16 to 0.61] 0.260
Age 0.01 [-0.01 to 0.02] 0.411
GenderMale 0.05 [-0.22 to 0.31] 0.741
ATB_in_last_6_monthsNO 0.24 [-0.14 to 0.62] 0.218
BDI_I_mildNo 0.02 [-0.35 to 0.40] 0.903
First_LanguageFrench/Luxembourgish/German -0.20 [-0.63 to 0.23] 0.357
Living_With_PartnerYes 0.31 [ 0.04 to 0.58] 0.026 ’*’
APOE4None 0.02 [-0.25 to 0.30] 0.882

With interaction

Outcome Model
Estimate [95% CI] P Value Sig
(Intercept) -4.10 [-7.24 to -1.11] 0.008 ’**’
Years_of_Education11-16 -1.33 [-2.18 to -0.48] 0.002 ’**’
Years_of_Education16+ -1.30 [-2.23 to -0.39] 0.005 ’**’
zInvSimpson -0.01 [-0.68 to 0.63] 0.975
Age 0.07 [ 0.03 to 0.11] <0.001 ’***’
GenderMale 1.08 [ 0.34 to 1.87] 0.006 ’**’
ATB_in_last_6_monthsNO -0.57 [-1.47 to 0.39] 0.226
BDI_I_mildNo -0.06 [-0.97 to 0.92] 0.899
First_LanguageFrench/Luxembourgish/German -0.32 [-1.36 to 0.82] 0.560
Living_With_PartnerYes -0.15 [-0.86 to 0.58] 0.683
APOE4None -0.66 [-1.36 to 0.05] 0.067
Years_of_Education11-16:zInvSimpson 0.19 [-0.62 to 1.02] 0.649
Years_of_Education16+:zInvSimpson -0.50 [-1.41 to 0.41] 0.282
Mediator Model
Estimate [95% CI] P Value Sig
(Intercept) -0.88 [-2.08 to 0.31] 0.147
Years_of_Education11-16 0.17 [-0.18 to 0.52] 0.342
Years_of_Education16+ 0.22 [-0.16 to 0.61] 0.260
Age 0.01 [-0.01 to 0.02] 0.411
GenderMale 0.05 [-0.22 to 0.31] 0.741
ATB_in_last_6_monthsNO 0.24 [-0.14 to 0.62] 0.218
BDI_I_mildNo 0.02 [-0.35 to 0.40] 0.903
First_LanguageFrench/Luxembourgish/German -0.20 [-0.63 to 0.23] 0.357
Living_With_PartnerYes 0.31 [ 0.04 to 0.58] 0.026 ’*’
APOE4None 0.02 [-0.25 to 0.30] 0.882

Note. Regression coefficients for mediator and outcome models, used for mediation analysis with alpha diversity measures. All analyses were adjusted for age, sex/gender, ATB, BDI-I, first language, PS and APOE. * P Value < .05, ** P Value < .01, *** P Value < .001.

Table 4. Mediation Analysis with Shannon Index as Mediator.

0-10 vs. 11-16

11-16 vs. 0-10 YEDU - Shannon - no interaction
Effect Estimate [95% CI] P Value Sig
Rcde 0.32 [ 0.15 to 0.69] 0.007 ’**’
Rpnde 0.32 [ 0.15 to 0.70] 0.007 ’**’
Rtnde 0.32 [ 0.15 to 0.70] 0.007 ’**’
Rpnie 0.96 [ 0.83 to 1.09] 0.526
Rtnie 0.96 [ 0.82 to 1.10] 0.526
Rte 0.31 [ 0.15 to 0.67] 0.004 ’**’
pm 0.02 [-0.06 to 0.17] 0.527

0-10 vs. 16+

16+ vs. 0-10 YEDU - Shannon - no interaction
Effect Estimate [95% CI] P Value Sig
Rcde 0.33 [ 0.14 to 0.75] 0.014 ’*’
Rpnde 0.33 [ 0.14 to 0.75] 0.014 ’*’
Rtnde 0.33 [ 0.14 to 0.75] 0.014 ’*’
Rpnie 0.96 [ 0.81 to 1.10] 0.531
Rtnie 0.96 [ 0.81 to 1.10] 0.531
Rte 0.31 [ 0.14 to 0.71] 0.008 ’**’
pm 0.02 [-0.05 to 0.20] 0.535

0-10 vs. 11-16 with interaction

11-16 vs. 0-10 YEDU - Shannon - with interaction
Effect Estimate [95% CI] P Value Sig
Rcde 0.31 [ 0.14 to 0.69] 0.007 ’**’
Rpnde 0.31 [ 0.14 to 0.69] 0.007 ’**’
Rtnde 0.31 [ 0.14 to 0.70] 0.008 ’**’
Rpnie 1.00 [ 0.80 to 1.25] 0.986
Rtnie 1.01 [ 0.79 to 1.40] 0.897
Rte 0.31 [ 0.15 to 0.67] 0.005 ’**’
ERcde -0.56 [-0.77 to -0.22] 0.007 ’**’
ERintref -0.13 [-0.26 to 0.03] 0.093
ERintmed 0.01 [-0.26 to 0.23] 0.961
ERpnie 0.00 [-0.20 to 0.25] 0.986
ERcde(prop) 0.81 [ 0.58 to 1.06] 0.003 ’**’
ERintref(prop) 0.19 [-0.07 to 0.46] 0.094
ERintmed(prop) -0.01 [-0.43 to 0.44] 0.963
ERpnie(prop) 0.00 [-0.42 to 0.35] 0.988
pm -0.01 [-0.21 to 0.17] 0.898
int 0.19 [-0.03 to 0.45] 0.070
pe 0.19 [-0.06 to 0.42] 0.107

0-10 vs. 16+ with interaction

16+ vs. 0-10 YEDU - Shannon - with interaction
Effect Estimate [95% CI] P Value Sig
Rcde 0.32 [ 0.13 to 0.74] 0.012 ’*’
Rpnde 0.36 [ 0.15 to 0.86] 0.022 ’*’
Rtnde 0.32 [ 0.14 to 0.74] 0.011 ’*’
Rpnie 1.00 [ 0.79 to 1.26] 0.985
Rtnie 0.88 [ 0.65 to 1.09] 0.260
Rte 0.32 [ 0.15 to 0.72] 0.009 ’**’
ERcde -0.55 [-0.78 to -0.18] 0.012 ’*’
ERintref -0.09 [-0.21 to 0.17] 0.344
ERintmed -0.04 [-0.37 to 0.17] 0.677
ERpnie 0.00 [-0.21 to 0.26] 0.985
ERcde(prop) 0.80 [ 0.56 to 1.05] 0.005 ’**’
ERintref(prop) 0.13 [-0.34 to 0.34] 0.336
ERintmed(prop) 0.06 [-0.28 to 0.71] 0.682
ERpnie(prop) 0.00 [-0.43 to 0.37] 0.988
pm 0.06 [-0.05 to 0.49] 0.266
int 0.19 [ 0.00 to 0.49] 0.053
pe 0.20 [-0.05 to 0.44] 0.089

Table 5. Mediation Analysis with Inverse Simpson as Mediator.

0-10 vs. 11-16

11-16 vs. 0-10 YEDU - InvSimpson - no interaction
Effect Estimate [95% CI] P Value Sig
Rcde 0.31 [ 0.15 to 0.67] 0.005 ’**’
Rpnde 0.31 [ 0.15 to 0.67] 0.005 ’**’
Rtnde 0.31 [ 0.15 to 0.67] 0.005 ’**’
Rpnie 0.99 [ 0.89 to 1.08] 0.832
Rtnie 0.99 [ 0.88 to 1.08] 0.832
Rte 0.31 [ 0.15 to 0.67] 0.005 ’**’
pm 0.01 [-0.05 to 0.09] 0.833

0-10 vs. 16+

16+ vs. 0-10 YEDU - InvSimpson - no interaction
Effect Estimate [95% CI] P Value Sig
Rcde 0.32 [ 0.14 to 0.73] 0.009 ’**’
Rpnde 0.32 [ 0.14 to 0.73] 0.009 ’**’
Rtnde 0.32 [ 0.14 to 0.73] 0.009 ’**’
Rpnie 0.98 [ 0.87 to 1.09] 0.762
Rtnie 0.98 [ 0.86 to 1.10] 0.762
Rte 0.31 [ 0.14 to 0.71] 0.008 ’**’
pm 0.01 [-0.05 to 0.12] 0.764

0-10 vs. 11-16 with interaction

11-16 vs. 0-10 YEDU - InvSimpson - with interaction
Effect Estimate [95% CI] P Value Sig
Rcde 0.31 [ 0.14 to 0.68] 0.005 ’**’
Rpnde 0.30 [ 0.15 to 0.66] 0.004 ’**’
Rtnde 0.31 [ 0.15 to 0.69] 0.006 ’**’
Rpnie 1.00 [ 0.85 to 1.20] 0.964
Rtnie 1.03 [ 0.89 to 1.24] 0.705
Rte 0.31 [ 0.15 to 0.67] 0.005 ’**’
ERcde -0.56 [-0.77 to -0.24] 0.005 ’**’
ERintref -0.13 [-0.25 to 0.00] 0.055
ERintmed 0.01 [-0.19 to 0.18] 0.966
ERpnie 0.00 [-0.15 to 0.20] 0.964
ERcde(prop) 0.82 [ 0.61 to 1.02] 0.004 ’**’
ERintref(prop) 0.20 [ 0.00 to 0.44] 0.054
ERintmed(prop) -0.01 [-0.33 to 0.33] 0.965
ERpnie(prop) 0.00 [-0.34 to 0.27] 0.965
pm -0.01 [-0.14 to 0.07] 0.705
int 0.18 [ 0.02 to 0.40] 0.039 ’*’
pe 0.18 [-0.02 to 0.39] 0.066

0-10 vs. 16+ with interaction

16+ vs. 0-10 YEDU - InvSimpson - with interaction
Effect Estimate [95% CI] P Value Sig
Rcde 0.32 [ 0.13 to 0.73] 0.010 ’*’
Rpnde 0.36 [ 0.16 to 0.83] 0.020 ’*’
Rtnde 0.32 [ 0.14 to 0.75] 0.010 ’*’
Rpnie 1.00 [ 0.83 to 1.25] 0.991
Rtnie 0.91 [ 0.67 to 1.10] 0.370
Rte 0.32 [ 0.15 to 0.73] 0.009 ’**’
ERcde -0.55 [-0.78 to -0.20] 0.010 ’*’
ERintref -0.09 [-0.19 to 0.17] 0.326
ERintmed -0.03 [-0.36 to 0.12] 0.701
ERpnie 0.00 [-0.17 to 0.25] 0.991
ERcde(prop) 0.81 [ 0.58 to 1.10] 0.003 ’**’
ERintref(prop) 0.14 [-0.35 to 0.33] 0.318
ERintmed(prop) 0.05 [-0.21 to 0.67] 0.705
ERpnie(prop) 0.00 [-0.40 to 0.31] 0.990
pm 0.05 [-0.07 to 0.41] 0.375
int 0.18 [-0.04 to 0.48] 0.084
pe 0.19 [-0.10 to 0.42] 0.114

Table 6. Taxonomic Classification of Identified Taxa.

Taxon Domain Phylum Class Order Family Genus
Bacilli Bacteria Firmicutes Bacilli
Lactobacillales Bacteria Firmicutes Bacilli Lactobacillales
Streptococcaceae Bacteria Firmicutes Bacilli Lactobacillales Streptococcaceae
Actinobacteria Bacteria Actinobacteriota Actinobacteria
Streptococcus Bacteria Firmicutes Bacilli Lactobacillales Streptococcaceae Streptococcus
Lachnospiraceae_UCG_001 Bacteria Firmicutes Clostridia Lachnospirales Lachnospiraceae Lachnospiraceae_UCG_001
OTU_000053 Bacteria Firmicutes Clostridia Oscillospirales Oscillospiraceae NK4A214_group
OTU_000508 Bacteria Firmicutes Clostridia Lachnospirales Lachnospiraceae Lachnospiraceae_UCG_001

Table 7. First Languages Spoken.

NC MCI
Danish 3 0
Dutch 2 1
English 7 0
French 27 5
German 20 3
Hungarian 1 0
Italian 2 2
Luxembourgish 135 44
Portuguese 1 3
Slovene 1 0
Spanish 1 0

Figure 1. Alpha Diversity Across Education Groups.

Note. Panels show results stratified by education groups with 0-10, 11-16 and 16+ years of education. Reported P values refer to Student’s t-Tests. InvSimpson = Inverse Simpson. Author MK.

Figure 2. Ordination Plots for MCI and Age Groups.

Ordination using Non-metric Multidimensional Scaling based on Bray-Curtis dissimilarity. P value (adonis2) adjusted for MCI, education, gender, age, ATB, BDI-I, first language, PS, APOE. Authors MK and VTEA.

A MCI

Df Sum Sq Mean Sq F value Pr(>F)
Groups 1 0.01 0.01 2.25 0.135
Residuals 256 0.63 0.00

B Age

Df Sum Sq Mean Sq F value Pr(>F)
Groups 2 0.00 0 0.01 0.987
Residuals 255 0.65 0

Figure 3. Relative Abundance Across Groups of Education.

Bacilli

Lactobacillales

Streptococcaceae

Actinobacteria

Streptococcus

Lachnospiraceae_UCG_001

OTU_000053

OTU_000508

End